Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down at the end of May, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: create a variable with estimated coefficients on dummies


From   Nick Cox <njcoxstata@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: create a variable with estimated coefficients on dummies
Date   Thu, 16 Aug 2012 10:00:10 +0100

Just to say sorry, but I don't understand this follow-up. The point
about using a dataset that everyone can access for examples is that
anyone who cares can see exactly and concretely what happens with
specific syntax. I don't understand what is implied by this discussion
with its counterfactuals about what the data might be.

At this moment, I am away from Stata 11 or 12 and cannot experiment to
work out the -merge- syntax needed, but I can't see that -reshape
long- is a wrong step.

Nick

On Thu, Aug 16, 2012 at 8:11 AM, Bernini, Michele
<Michele.Bernini@unitn.it> wrote:
> Sorry for this follow-up. Nick's method resolves the problem except for the reshape part. Take the auto example. If the variable 'rep78' had three categories A, B, C, the command i.rep78 would have created three dummies and after the  command  "xi : statsby , by(foreign) : regress mpg i.rep78" I would have three columns :  "_b_Ir~78_A" , " _b_Ir~78_B", " _b_Ir~78_C". When I reshape long I would need A, B, C instead of "foreign" and "domestic" to be my the values assumed by my id variable (even if I need to run separate regressions for foreign and domestic).
>
> Thanks in advance for your advices,
>
> Michele
>
>
> On 16 Aug 2012, at 06:31, Bernini, Michele wrote:
>
>> Thanks a lot Nick for your suggestion. It works very well.
>> On 15 Aug 2012, at 20:14, Nick Cox wrote:
>>
>>> I wouldn't approach it this way. You can save the coefficients to a
>>> different dataset
>>>
>>> sysuse auto, clear
>>> xi : statsby , by(foreign) : regress mpg i.rep78
>>> reshape long _b_Irep78_ , i(foreign)
>>>
>>> . l
>>>
>>>    +--------------------------------------+
>>>    |  foreign   _j   _b_Ir~78_    _b_cons |
>>>    |--------------------------------------|
>>> 1. | Domestic    2      -1.875         21 |
>>> 2. | Domestic    3          -2         21 |
>>> 3. | Domestic    4   -2.555556         21 |
>>> 4. | Domestic    5          11         21 |
>>> 5. |  Foreign    2           0   23.33333 |
>>>    |--------------------------------------|
>>> 6. |  Foreign    3           0   23.33333 |
>>> 7. |  Foreign    4    1.555556   23.33333 |
>>> 8. |  Foreign    5           3   23.33333 |
>>>    +--------------------------------------+
>>>
>>> and then (if I understand what you want) -merge- back to the original dataset.
>>>
>>> On Wed, Aug 15, 2012 at 7:27 PM, Bernini, Michele
>>> <Michele.Bernini@unitn.it> wrote:
>>>> Dear Statalister,
>>>>
>>>> I am programming a loop to execute  the following operations:
>>>>
>>>> 1- estimate a regression on a big set of dummies (country-hs6 product categories)
>>>>
>>>> 2-create a variable with the estimated coefficients on the dummies. For each observation (row) the value of this variable should correspond to the coeff. on the dummy that assumes value 1 for that observation.
>>>>
>>>> I tried something like that:
>>>>
>>>> gen ch=.
>>>> levelsof hs4, local(lvl_hs4) \\ level at which I run separate regressions
>>>> foreach element in `lvl_hs4'{
>>>> xi: reg Y X i.CH i.t if hs4==`element'
>>>> levelsof CH, local(clist)
>>>> foreach c of local clist {
>>>> replace ch = _b[_ICH_`c'] if CH == `c' \\ here I try to give the variable ch the value of the coefficients
>>>> }
>>>> }
>>>>
>>>> my problem is that the local `clist' includes also the missing category of the dummies, so when it looks for _b[_ICH_`c'] where c = "missing category" the loop aborts. In either words I would either need the local `clist' to include only the dummies created from CH in each regression, or to find another way to do that. I also tried with svmat but then I cannot match coefficients on the new variable 'ch' with the observations with each category in the dummies.

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index